A systematic method for fault diagnosis of analogue circuits based on the combination of neural networks and wavelet transforms is presented. Using wavelet decomposition as a tool for removing noise from the sampled signals, optimal feature information is extracted by wavelet noise removal, multi-resolution decomposition, PCA (principal component analysis) and data normalisation. The features are applied to the proposed wavelet neural network and the fault patterns are classified. Diagnosis principles and procedures are described. The reliability of the method and comparison with other methods are shown by two active filter examples.